Add verify_new_header

[yihuang]

Add verify_new_header to combine the verifications of lite client logic when syncing new header.
Changed TrustedState to trait.

[ebuchman]

Thanks for this, but take a look at the new approach in #114.

Note sure the Error -> VerifyError is helpful since Verify prefix doesn't really tell us much. We should make some improvements to the error systems though - we've started tracking some ideas in #110 . The impl Display from this PR would be good to preserve too if you want to submit that in another PR. Will close for now, but let us know if you think something else from this should be included.

[yihuang]

The name Error is a conflict with std::error::Error, the Display trait is a requirement of Error trait, that's it.
I'll look at the new approach, thanks.

[tarcieri]

To avoid Error conflicting with std::error::Error, you can refer to the latter using its fully qualified name.

If you use a crate like thiserror to impl Display and std::error::Error using a proc macro, you don't even have to worry about referring to std::error::Error directly, as it will use the proc macro to write the impl for you.

[yihuang]

To avoid Error conflicting with std::error::Error, you can refer to the latter using its fully qualified name.

If you use a crate like thiserror to impl Display and std::error::Error using a proc macro, you don't even have to worry about referring to std::error::Error directly, as it will use the proc macro to write the impl for you.

If you wrap it inside your own Error type, then you can implement whatever trait your need, but if you use the tendermint-rs one directly and hoping it has some trait which is not implemented, there's nothing you can do. I just think there is no reason the Error type is not an instance of std::error::Error.

[tarcieri]

It definitely should impl std::error::Error. I'm just saying you don't need to change its name to do that.